HW 06 Sol

Name: SOLUTION (Havlicek)
Section:
Laboratory Exercise 3
DISCRETE-TIME SIGNALS: FREQUENCY-DOMAIN
REPRESENTATIONS
3.1
DISCRETE-TIME FOURIER TRANSFORM
Project 3.1
DTFT Computation
A copy of Program P3_1 is given below:

% Program P3_1
% Evaluation of the DTFT
clf;
% Compute the frequency samples of the DTFT
w = -4*pi:8*pi/511:4*pi;
num = [2 1];den = [1 -0.6];
h = freqz(num, den, w);
% Plot the DTFT
subplot(2,1,1)
plot(w/pi,real(h));grid
title('Real part of H(e^{j\omega})')
xlabel('\omega /\pi');
ylabel('Amplitude');
subplot(2,1,2)
plot(w/pi,imag(h));grid
title('Imaginary part of H(e^{j\omega})')
pause
subplot(2,1,1)
plot(w/pi,abs(h));grid
title('Magnitude Spectrum |H(e^{j\omega})|')
subplot(2,1,2)
plot(w/pi,angle(h));grid
title('Phase Spectrum arg[H(e^{j\omega})]')
ylabel('Phase in radians');
Answers:
Q3.1
The expression of the DTFT being evaluated in Program P3_1 is - H e
2 z 1
1 0.6 z 1
The function of the pause command is - to pause execution of a Matlab program. Without
arguments, pause waits for the user to type any key. With an argument, pause pauses
for a number of seconds specified by the argument.
The plots generated by running Program P3_1 are shown below:

Real part of H(ej )
Amplitude
8
6
4
2
0
-4
-3
-2
-1
0
/
Imaginary part of H(ej )
Amplitude
4
2
0
-2
-4
-4
-3
-2
-1
Magnitude Spectrum |H(ej )|
Amplitude
8
6
4
2
0
-4
-3
-2
-1
0
/
Phase Spectrum arg[H(ej )]

2
Phase in radians
Q3.2
1
0
-1
-2
-4
-3
-2
-1
The DTFT is a _periodic_ function of .

Its period is - 2
The types of symmetries exhibited by the four plots are as follows:
Q3.3
The real part is 2 periodic and EVEN SYMMETRIC.
The imaginary part is 2 periodic and ODD SYMMETRIC.
The magnitude is 2 periodic and EVEN SYMMETRIC.
The phase is 2 periodic and ODD SYMMETRIC.
The required modifications to Program P3_1 to evaluate the given DTFT of Q3.3 are given below:
% Program P3_1B
clf;
% because 0 \leq w \leq pi is the default for "freqz",
% the vector "w" is now an output of freqz instead of an input.
N = 512;
num = [0.7 -0.5 0.3 1];
den = [1 0.3 -0.5 0.7];
[h,w] = freqz(num, den, N);
% Plot the DTFT
subplot(2,1,1)
subplot(2,1,2)
pause
subplot(2,1,1)
subplot(2,1,2)
The plots generated by running the modified Program P3_1 are shown below:
Real part of H(ej )
Amplitude
1
0.5
0
-0.5
-1
0.1
0.2
0.3
0.4
0.5
/
0.6
0.7
0.8
0.9
0.7
0.8
0.9
0.8
0.9
0.8
0.9
Amplitude
1
0.5
0
-0.5
-1
0.1
0.2
0.3
0.4
0.5
0.6
Amplitude
1
1
1
0.1
0.2
0.3
0.4
0.5
/
0.6
0.7

Phase in radians
4
2
0
-2
-4
0.1
0.2
0.3
0.4
0.5
The DTFT is a _periodic__ function of .
0.6
0.7
Its period is - 2
The jump in the phase spectrum is caused by - a branch cut in the arctan function used by
angle in computing the phase. angle returns the principal branch of arctan.
The phase spectrum evaluated with the jump removed by the command unwrap is as given
below:
0
-1
Phase in radians
-2
-3
-4
-5
-6
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Q3.4
The required modifications to Program P3_1 to evaluate the given DTFT of Q3.4 are given below:
% Program P3_1D
clf;
w = -4*pi:8*pi/511:4*pi;
num = [1 3 5 7 9 11 13 15 17];
den = 1;
h = freqz(num, den, w);
% Plot the DTFT
subplot(2,1,1)
subplot(2,1,2)
pause
subplot(2,1,1)
subplot(2,1,2)
The plots generated by running the modified Program P3_1 are shown below:
Real part of H(ej )
Amplitude
100
50
0
-50
-4
-3
-2
-1
0
/
Amplitude
100
50
0
-50
-100
-4
-3
-2
-1
Amplitude
100
50
0
-4
-3
-2
-1
/
Phase in radians
4
2
0
-2
-4
-4
-3
-2
-1
The DTFT is a _periodic_ function of .

Its period is - 2
The jump in the phase spectrum is caused by - angle returns the principal value of the
arc tangent.
Q3.5
The required modifications to Program P3_1 to plot the phase in degrees are indicated below:
Only the last paragraph of the code needs to be changed to:

% plot phase in degrees
subplot(2,1,2)
plot(w/pi,180*angle(h)/pi);grid
ylabel('Phase in degrees');
Project 3.2
DTFT Properties
Answers:
Q3.6
The modified Program P3_2 created by adding appropriate comment statements, and adding
program statements for labeling the two axes of each plot being generated by the program is
given below:
% Program P3_2B
% Time-Shifting Properties of DTFT
clf;
w = -pi:2*pi/255:pi; % freqency vector for evaluating DTFT
D = 10; % Amount of time shift in samples
num = [1 2 3 4 5 6 7 8 9];
% h1 is the DTFT of original sequence
% h2 is the DTFT of the time shifted sequence
h1 = freqz(num, 1, w);
h2 = freqz([zeros(1,D) num], 1, w);
subplot(2,2,1)
% plot the DTFT magnitude of the original sequence
plot(w/pi,abs(h1));grid
title('Magnitude Spectrum of Original Sequence','FontSize',8)
% plot the DTFT magnitude of the shifted sequence
subplot(2,2,2)
title('Magnitude Spectrum of Time-Shifted Sequence','FontSize',8)
% plot the DTFT phase of the original sequence
subplot(2,2,3)
plot(w/pi,angle(h1));grid
title('Phase Spectrum of Original Sequence','FontSize',8)
% plot the DTFT phase of the shifted sequence
subplot(2,2,4)
title('Phase Spectrum of Time-Shifted Sequence','FontSize',8)
The parameter controlling the amount of time-shift is - D
Q3.7
The plots generated by running the modified program are given below:
Magnitude Spectrum of Time-Shifted Sequence
60
40
40
Amplitude
Amplitude
Magnitude Spectrum of Original Sequence
60
20
0
-1
-0.5
0
/
0.5
0
-1
Phase Spectrum of Original Sequence
0
/
0.5
4
Phase in radians
Phase in radians
-0.5
Phase Spectrum of Time-Shifted Sequence
4
2
0
-2
-4
-1
20
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
From these plots we make the following observations: The time shift does not have any
effect at all on the magnitude spectrum. However, it has a significant effect on the
phase spectrum. The effect is to add phase, which makes the slope of the phase
function steeper (more negative in this case).
Q3.8
Program P3_2 was run for the following value of the time-shift D=5.
60
40
40
Amplitude
Amplitude
60
20
0
-1
-0.5
0
/
0.5
0
-1
0.5
4
Phase in radians
Phase in radians
-0.5
4
2
0
-2
-4
-1
20
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
From these plots we make the following observations: As before, of course, the time shift
has no effect on the magnitude spectrum. However, there is a very significant change
to the phase. As before, the time shift adds phase to the DTFT, making the slope of the
phase spectrum steeper. However, in this case with D=5 instead of D=10, the
(negative) increase in the slope is less than it was before.
Q3.9
Program P3_2 was run for the following values of the time-shift and for the following values of
length for the sequence
1. Length 4, time shift D=3.

2. Length 16, time shift D=12.
10
10
8
Amplitude
Amplitude
10
6
4
2
-1
-0.5
0
/
0.5
6
4
2
-1
Phase in radians
Phase in radians
0.5
2
0
-2
-4
-1
-0.5
0
/
0.5
2
0
-2
-4
-1
100
100
Amplitude
150
50
-0.5
0
/
0.5
0.5
50
0
-1
-0.5
0.5
4
Phase in radians
4
2
0
-2
-4
-1
150
0
-1
-0.5
Amplitude
Phase in radians
-0.5
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
From these plots we make the following observations: Increasing the length makes the
magnitude spectrum more narrow (i.e., makes the signal more low pass). It also
makes the phase steeper (i.e., the slope more negative). This is because, if we think of
the sequence as the impulse response of an LTI system, increasing the length adds more
11
delay between the input and output of the system. As before, the time shift does not
have any effect on the magnitude spectrum. However, it makes the slope of the phase
spectrum more negative. The larger the value of the delay, the more negative slope is
added to the phase spectrum.
Q3.10
given below:
% Program P3_3B
% Frequency-Shifting Properties of DTFT
clf;
wo = 0.4*pi; % Amount of frequency shift in radians
% h1 is the DTFT of the original sequence
% h2 is the DTFT of the frequency shifted sequence
num1 = [1 3 5 7 9 11 13 15 17];
L = length(num1);
h1 = freqz(num1, 1, w);
n = 0:L-1;
num2 = exp(wo*i*n).*num1;
h2 = freqz(num2, 1, w);
% plot the DTFT magnitude of the original sequence
subplot(2,2,1)
% plot the DTFT magnitude of the freq shifted sequence
subplot(2,2,2)
title('Magnitude Spectrum of Frequency-Shifted Sequence','FontSize',8)
% plot the DTFT phase of the original sequence
subplot(2,2,3)
% plot the DTFT phase of the shifted sequence
subplot(2,2,4)
title('Phase Spectrum of Frequency-Shifted Sequence','FontSize',8)
The parameter controlling the amount of frequency-shift is - wo.

Q3.11
12
Magnitude Spectrum of Frequency-Shifted Sequence
100
Amplitude
Amplitude
100
50
0
-1
-0.5
0
/
0.5
50
0
-1
0.5
Phase Spectrum of Frequency-Shifted Sequence
4
Phase in radians
Phase in radians
4
2
0
-2
-4
-1
-0.5
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
From these plots we make the following observations:
Both the magnitude and phase

spectra are shifted right by wo, which is given by 0.4 in this case. Note that the
frequency shifted signal was obtained by multiplying the original sequence pointwise
with a complex-valued exponential sequence. Thus, the frequency shifted sequence is
also complex-valued and its DTFT does not have conjugate symmetry.
Q3.12
Program P3_3 was run for the following value of the frequency-shift
13
wo = -0.5.
100
Amplitude
Amplitude
100
50
0
-1
-0.5
0
/
0.5
50
0
-1
0.5
4
Phase in radians
Phase in radians
4
2
0
-2
-4
-1
-0.5
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
In this case, the magnitude and

phase spectra are shifted left by /2 rad. As before, the frequency shifted signal is
complex-valued, so the frequency shift causes a loss of the conjugate symmetry that
was present in the spectrum of the original signal. NOTE: you should keep in mind
that these spectra are all 2-periodic; we are only displaying the fundamental period.
Q3.13
Program P3_3 was run for the following values of the frequency-shift and for the following
values of length for the sequence
1. Length 4, frequency shift wo = .

2. Length 16, frequency shift wo = -0.3.
14
20
15
15
Amplitude
Amplitude
20
10
5
0
-1
-0.5
0
/
0.5
10
5
0
-1
Phase in radians
Phase in radians
0.5
2
0
-2
-4
-1
-0.5
0
/
0.5
2
0
-2
-4
-1
200
200
Amplitude
300
100
-0.5
0
/
0.5
0.5
100
0
-1
-0.5
0.5
4
Phase in radians
4
2
0
-2
-4
-1
300
0
-1
-0.5
Amplitude
Phase in radians
-0.5
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
From these plots we make the following observations: The original sequences have a low
pass characteristic. As before with the time shift property, a shorter length gives a
broader low pass magnitude spectrum, whereas a longer length results in a low pass
magnitude spectrum that is more narrow. Also, if we consider these sequences to be
15
the impulse responses of LTI systems, then a shorter length implies a smaller delay
between the input and output of the system, which corresponds to a phase spectrum
with a negative slope that is less steep. Likewise, a longer length implies a longer
delay between the input and output, which corresponds to a phase spectrum with a
negative slope that is steeper. In both cases shown here, the effect of the frequency
shift is to translate both the magnitude and phase spectrum by wo radians. Whereas the
original sequences are real-valued and hence have conjugate symmetric spectra, the
frequency shifted sequences are complex-valued and do not exhibit any inherent
spectral symmetry.
Q3.14
given below:
% Program P3_4B
% Convolution Property of DTFT
clf;
x1 = [1 3 5 7 9 11 13 15 17]; % first sequence
x2 = [1 -2 3 -2 1]; % second sequence
y = conv(x1,x2); % time domain convolution of x1 and x2
h1 = freqz(x1, 1, w); % DTFT of sequence x1
% hp is the pointwise product of the two DTFT's
hp = h1.*h2;
% h3 is the DTFT of the time domain convolution;
%
it should be the same as hp
h3 = freqz(y,1,w);
% plot the magnitude of the product of the two original spectra
subplot(2,2,1)
plot(w/pi,abs(hp));grid
title('Product of Magnitude Spectra','FontSize',8)
% plot the magnitude spectrum of the time domain convolution
subplot(2,2,2)
title('Magnitude Spectrum of Convolved Sequence','FontSize',8)
% plot the phase of the product of the two original spectra
subplot(2,2,3)
plot(w/pi,angle(hp));grid
title('Sum of Phase Spectra','FontSize',8)
% plot the phase spectrum of the time domain convolution
subplot(2,2,4)
title('Phase Spectrum of Convolved Sequence','FontSize',8)
16
Q3.15
Product of Magnitude Spectra
Magnitude Spectrum of Convolved Sequence
100
Amplitude
Amplitude
100
50
0
-1
-0.5
0
/
0.5
50
0
-1
Sum of Phase Spectra
0.5
4
Phase in radians
Phase in radians
0
/
Phase Spectrum of Convolved Sequence
4
2
0
-2
-4
-1
-0.5
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
From these plots we make the following observations: The DTFT magnitude and phase
spectra obtained by performing pointwise multiplication of the two DTFTs of the

original sequences are identical to those obtained by performing time domain
convolution of the two original sequences; this verifies the convolution property of the
DTFT.
Q3.16
Program P3_4 was run for the following two different sets of sequences of varying lengths
1. Length of x1 = 8; x1[n] = 1 2 for 0 n 7;

Length of x2 = 4; x2[n]= [0.25 0.25 0.25 0.25]
n

Length of x2 = 8; x2[n]= [1 3 5 7 9 11 13 15]
n
17
1.5
1.5
Amplitude
Amplitude
1
0.5
0
-1
-0.5
0
/
0.5
Phase in radians
Phase in radians
-2
-0.5
0
/
0.5
2
0
-2
-4
-1
30
30
Amplitude
40
20
10
0
/
0.5
40
-0.5
-0.5
Amplitude
0.5
0.5
20
10
0
-1
-0.5
0.5
4
Phase in radians
4
2
0
-2
-4
-1
0
-1
-0.5
Phase in radians
0.5
0
-1
-4
-1
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
The convolution property of the

DTFT is again verified in both cases. In each case, the DTFT magnitude and phase
obtained by taking the pointwise products of the DTFTs of the two original sequences
18
are identical to the magnitude and phase spectra obtained from the DTFT of the time
domain convolution of the two original sequences.
Q3.17
given below:
% Program P3_5B
% Modulation Property of DTFT
clf;
x1 = [1 3 5 7 9 11 13 15 17]; % first sequence
x2 = [1 -1 1 -1 1 -1 1 -1 1]; % second sequence
% y is the time domain pointwise product of x1 and x2
y = x1.*x2;
h3 = freqz(y,1,w);
% DTFT of sequence y
% plot the magnitude spectrum of x1
subplot(3,1,1)
title('Magnitude Spectrum of First Sequence')
% plot the magnitude spectrum of x2
subplot(3,1,2)
title('Magnitude Spectrum of Second Sequence')
% plot the magnitude spectrum of y
%
it should be 1/2pi times the convolution of the DTFT's
%
of the two original sequences.
subplot(3,1,3)
title('Magnitude Spectrum of Product Sequence')
Q3.18
19
Amplitude
Magnitude Spectrum of First Sequence

100
50
0
-1
-0.8
-0.6
-0.4
-0.2
0.2
0.4
0.6
0.8
0.8
0.8
/
Amplitude
Magnitude Spectrum of Second Sequence

10
5
0
-1
-0.8
-0.6
-0.4
-0.2
0.2
0.4
0.6
/
Amplitude
Magnitude Spectrum of Product Sequence

100
50
0
-1
-0.8
-0.6
-0.4
-0.2
0
/
0.2
0.4
0.6
The DTFT of the product sequence

y is 1/2 times the convolution of the DTFTs of the two sequences x1 and x2, as
expected. The low-pass mainlobe of the DTFT of x1 combines with the high-pass
mainlobe of the DTFT of x2 to produce a high-pass mainlobe centered at in the
magnitude spectrum of the product signal. The low-pass mainlobe of the DTFT of x1
combines with the low-pass sidelobes of the DTFT of x2 to produce a low-pass smooth
region of relatively lower gain centered at DC in the magnitude spectrum of the product
signal.
Q3.19

Length of x2 = 8; x2[n]= [1/8 1/8 1/8 1/8 1/8 1/8 1/8 1/8]
n

Length of x2 = 16; x2[n]= [1 3 5 7 9 11 13 15 0 0 0 0 0 0 0 0]
n
20
Amplitude

2
1
0
-1
-0.8
-0.6
-0.4
-0.2
0.2
0.4
0.6
0.8
0.8
0.6
0.8
0.6
0.8
0.8
0.8
/
Amplitude

1
0.5
0
-1
-0.8
-0.6
-0.4
-0.2
0.2
0.4
0.6
/
Amplitude

0.4
0.2
0
-1
-0.8
-0.6
-0.4
-0.2
0
/
0.2
0.4
Amplitude

4
2
0
-1
-0.8
-0.6
-0.4
-0.2
0.2
0.4
/
Amplitude

100
50
0
-1
-0.8
-0.6
-0.4
-0.2
0.2
0.4
0.6
/
Amplitude

20
10
0
-1
-0.8
-0.6
-0.4
-0.2
0
/
0.2
0.4
0.6
In the first example, both x1 and x2

are low pass sequences. Moreover, the DTFT of x2 is a sync pulse. Taking the
21
product of these two sequences produces a new sequence y that is also low pass in
character. The magnitude spectrum of y has a shape that is very similar to that of x1,
but with some averaging. The spectral magnitude of y is reduced compared to that of
x1, primarily due to the division by 2 that is inherent in the DTFT frequency
convolution (time modulation) property. In the second example, x1 is high pass while
x2 is low pass. As before in Q3.18, we see that magnitude spectrum of y is essentially
high pass in character as a result of the low pass mainlobe of the DTFT of x2 being
convolved with the high pass mainlobe of x1.
Q3.20
given below:
% Program P3_6B
% Time Reversal Property of DTFT
clf;
% original ramp sequence
% note: num is nonzero for 0 <= n <= 3.
num = [1 2 3 4];
L = length(num)-1;
h1 = freqz(num, 1, w); % DTFT of original ramp sequence
% h2 contains the sample values of h1 in reverse order, but
%
it is NOT the time reversed version of h1. The time
%
reversed version must be nonzero for -3 <= n <= 0. However,
%
h2 is nonzero for 0 <= n <= 3. So, to get the time reversed
%
version of h1, we still need to time SHIFT h2 to the left.
%
This is accomplished in the frequency domain using the time
%
shift property of the DTFT. Thus, h3, which IS the time
%
reversed version of h1, is obtained by multiplying h2 times
%
a linear phase term to accomplish the required time shift.
h2 = freqz(fliplr(num), 1, w);
h3 = exp(w*L*i).*h2;
% plot the magnitude spectrum of the original ramp sequence
subplot(2,2,1)
% plot the magnitude spectrum of the time reversed ramp sequence
subplot(2,2,2)
title('Magnitude Spectrum of Time-Reversed Sequence','FontSize',8)
% plot the phase spectrum of the original ramp sequence
subplot(2,2,3)
% plot the phase spectrum of the time reversed ramp sequence
subplot(2,2,4)
title('Phase Spectrum of Time-Reversed Sequence','FontSize',8)
22
The original ramp

sequence is nonzero for 0 n 3. A new sequence is formed by using fliplr; this
new sequence contains the samples of the original ramp sequence in time reversed
order. However, the new sequence is still nonzero for 0 n 3, whereas the time
reversed ramp sequence must be nonzero for -3 n 0. This required left shift in time
is accomplished in the frequency domain using the time shift property of the DTFT as
follows. First, freqz is called to set h2 equal to the DTFT of the new sequence
obtained from calling fliplr on the original ramp sequence. Finally, h3 is set equal
to the DTFT of the time reversed ramp by multiplying h2 times a linear phase term to
implement the required left shift in the time domain.
The program implements the time-reversal operation as follows
Magnitude Spectrum of Time-Reversed Sequence
10
10
8
Amplitude
Amplitude
6
4
2
-1
-0.5
0
/
0.5
-0.5
0.5
Phase Spectrum of Time-Reversed Sequence
Phase in radians
2
0
-2
-4
-1
2
-1
4
Phase in radians
Q3.21
-0.5
0
/
0.5
2
0
-2
-4
-1
-0.5
0.5
Both the original and time reversed

ramp sequences are real-valued. Therefore, they have conjugate symmetric DTFTs.
For both sequences, this implies that the magnitude spectrum is even symmetric and the
phase spectrum is odd symmetric. Now, the DTFT of the time reversed ramp (h3) is
equal to a frequency reversed version of the DTFT of the original sequence (h1). Since
flipping an even function has no net effect, we see in the graphs above that both the
original and time reversed sequences have identical magnitude spectra. However, since
the phase spectra are odd symmetric, the phase spectrum of the time reversed sequence
is a frequency reversed version of the phase spectrum of the original sequence.
23
1. Length of num = 8; num = 1 2 for 0 n 7

n
2. Length of num = 16; num = 3 4 for 0 n 15

n
1.5
1.5
Amplitude
Amplitude
0.5
-1
-0.5
0
/
0.5
-0.5
0.5
Phase in radians
0.5
0
-0.5
-1
-1
0.5
-1
1
Phase in radians
Q3.22
-0.5
0
/
0.5
0.5
0
-0.5
-1
-1
24
-0.5
0
/
0.5
3
Amplitude
Amplitude
2
1
0
-1
-0.5
0
/
0.5
-0.5
0.5
1
Phase in radians
Phase in radians
1
0
-1
1
0.5
0
-0.5
-1
-1
-0.5
0
/
0.5
0.5
0
-0.5
-1
-1
-0.5
0.5
As before in Q3.20, we see that the

DTFT of the time reversed sequence is a frequency reversed version of the DTFT of the
original sequence. In particular, because these are real-valued sequences with even
magnitude spectra and odd phase spectra, the magnitude spectra of the original and
time reversed sequences are the same. However, the phase spectrum of the time
reversed sequence is a frequency reversed version of the phase spectrum of the original
sequence.
25
3.2
DISCRETE FOURIER TRANSFORM
Project 3.3
DFT and IDFT Computations
Answers:
Q3.23
The MATLAB program to compute and plot the L-point DFT X[k] of a length-N sequence
x[n] with L N and then to compute and plot the IDFT of X[k] is given below:
% Program P3_3DFT
% Compute and plot the L-point DFT of an N-point signal, L >= N.
clf;
%Initialize
N=200; % length of signal
L=256; % length of DFT
nn = [0:N-1];
kk = [0:L-1];
% the signal x
xR = [0.1*(1:100) zeros(1,N-100)]; % real part
xI = [zeros(1,N)]; % imag part
x = xR + i*xI;
% DFT
XF = fft(x,L);
% plot xR and xI
subplot(3,2,1);grid;
plot(nn,xR);grid;
title('Re\{x[n]\}');
xlabel('Time index n');
subplot(3,2,2);
plot(nn,xI);grid;
title('Im\{x[n]\}');
% plot real and imag parts of DFT
subplot(3,2,3);
plot(kk,real(XF));grid;
title('Re\{X[k]\}');
xlabel('Frequency index k');
subplot(3,2,4);
plot(kk,imag(XF));grid;
title('Im\{X[k]\}');
% IDFT
xx = ifft(XF,L);
% plot real and imaginary parts of the IDFT
subplot(3,2,5);
plot(kk,real(xx));grid;
title('Real part of IDFT\{X[k]\}');
subplot(3,2,6);
plot(kk,imag(xx));grid;
title('Imag part of IDFT\{X[k]\}');
26
The DFT and the IDFT pairs generated by running the program for sequences of different
lengths N and for different values of the DFT length L are shown below:
Im{x[n]}
Amplitude
5
0
50
100
150
Time index n
Re{X[k]}
200
Amplitude
1000
0
0
100
200
Frequency index k
Real part of IDFT{X[k]}
300
10
0
-10
100
200
Time index n
1
0
-1
50
100
150
Time index n
Im{X[k]}
200
100
200
Frequency index k
Imag part of IDFT{X[k]}
300
100
200
Time index n
300
500
0
-500
Amplitude
Amplitude
-1000
300
1
0
-1
Amplitude
Im{x[n]}
Amplitude
1
0
-1
100
200
Time index n
Re{X[k]}
300
Amplitude
50
0
-50
100
200
Frequency index k
300
2
0
-2
100
200
Time index n
300
27
1
0
-1
100
200
Time index n
Im{X[k]}
300
100
200
Frequency index k
300
100
200
Time index n
300
100
0
-100
Amplitude
Amplitude
Re{x[n]}
Amplitude
Amplitude
Amplitude
Re{x[n]}
10
1
0
-1
Im{x[n]}
Amplitude
0
-1
100
200
Time index n
Re{X[k]}
300
Amplitude
0
-200
100
200
Frequency index k
300
Amplitude
1
0
-1
100
200
Time index n
1
0
-1
100
200
Time index n
Im{X[k]}
300
100
200
Frequency index k
300
100
200
Time index n
300
-13
200
Amplitude
Amplitude
Amplitude
Re{x[n]}
1
300
x 10
0
-2
1
0
-1
From these plots we make the following observations: In the first example, the signal is a
real-valued ramp of length 100 zero padded on the right for a total signal length of
N=200. The length of the DFT is L=256. We see that the 256-point DFT is conjugate
symmetric as expected. The signal obtained from the IDFT has a length of L=256, but
is otherwise identical to the original signal up to roundoff.
The second example is a real-valued cosine with frequency = /16 and length N=256.
The length of the DFT is L=275. The signal would be a sum of two DFT basis
functions if N and L were equal. However, since this is not the case, we see that there
are more than two nonzero samples in the DFT. Moreover, since the zero padded Lpoint version of the signal does not have even symmetry, there are nonzero frequency
samples in both the real and imaginary parts of the DFT.
In the third example, the signal is the same as in the second example. However, in this
case we have N=L. So there is no zero padding and the signal that is passed to the FFT
routine is real and even. Consequently, the DFT is also real and even up to roundoff.
Moreover, in this case the signal is a simple sum of two DFT basis functions.
Therefore, we see only two nonzero samples in the DFT array (to within roundoff).
28
Q3.24
The MATLAB program to compute the N-point DFT of two length-N real sequences using a
single N-point DFT and compare the result by computing directly the two N-point DFTs is given
below:
% Program Q3_24
% Use a single N-point DFT to compute simultaneously the N-point
%
DFT's of two real-valued N-point sequences.
clf;
%Initialize
N=256; % length of signal
nn = [0:N-1];
ntime = [-N/2:N/2-1];
g = (0.75).^abs(ntime); % signal g
h = (-0.9).^ntime; % signal h
% DFT's of g and h
GF = fft(g);
HF = fft(h);
x = g + i*h; % the composite signal x
% DFT of composite signal
XF = fft(x);
% DFT of g derived from composite DFT XF
XFstar = conj(XF);
XFstarmod = [XFstar(1) fliplr(XFstar(2:N))];
GF2 = 0.5*(XF + XFstarmod);
HF2 = -i*0.5*(XF - XFstarmod);
abs(max(GF-GF2))
abs(max(HF-HF2))
% plot real and imag parts of direct computation of GF
figure(1);clf;
plot(nn,real(GF));grid;
title('Two N-point DFT''s');
ylabel('Re\{G[k]\}');
subplot(2,2,2);
plot(nn,imag(GF));grid;
ylabel('Im\{G[k]\}');
% plot real and imag parts of composite computation of GF
plot(nn,real(GF2));grid;
title('Single N-point DFT');
ylabel('Re\{G[k]\}');
subplot(2,2,4);
plot(nn,imag(GF2));grid;
ylabel('Im\{G[k]\}');
% plot real and imag parts of direct computation of HF
figure(2);clf;
plot(nn,real(HF));grid;
xlabel('Freq index k');
29
ylabel('Re\{H[k]\}');
subplot(2,2,2);
plot(nn,imag(HF));grid;
ylabel('Im\{H[k]\}');
% plot real and imag parts of composite computation of HF
plot(nn,real(HF2));grid;
ylabel('Re\{H[k]\}');
subplot(2,2,4);
plot(nn,imag(HF2));grid;
ylabel('Im\{H[k]\}');
The DFTs generated by running the program for sequences of different lengths N are shown
below:
-16
Two N-point DFT's

10
10
Im{G[k]}
Re{G[k]}
5
0
-5
-10
100
200
Frequency index k
Single N-point DFT
10
Two N-point DFT's
-5
300
x 10
100
200
Frequency index k
-10
x 10 Single N-point DFT
300
300
Im{G[k]}
Re{G[k]}
5
0
-5
-10
100
200
Frequency index k
-5
300
30
100
200
Frequency index k
x 10
Two N-point DFT's

4
4
2
0
100
200
Freq index k
6
-4
300
100
200
Freq index k
6
300
300
2
Im{H[k]}
Re{H[k]}
0
-2
6
4
2
0
0
-2
100
200
Freq index k
-4
300
Two N-point DFT's
Im{G[k]}
Re{G[k]}
50
100
Frequency index k
Single N-point DFT
50
100
Frequency index k
Single N-point DFT
150
50
100
Frequency index k
150
Im{G[k]}
Re{G[k]}
-5
150
-5
100
200
Freq index k
Two N-point DFT's
-5
Two N-point DFT's
2
Im{H[k]}
Re{H[k]}
x 10
50
100
Frequency index k
-5
150
31
Two N-point DFT's
Two N-point DFT's
40
40
20
Im{H[k]}
Re{H[k]}
20
-20
-20
50
100
Freq index k
Single N-point DFT
-40
150
40
50
100
Freq index k
Single N-point DFT
150
50
100
Freq index k
150
20
Im{H[k]}
Re{H[k]}
40
20
-20
0
-20
50
100
Freq index k
-40
150
Two N-point DFT's

2
1
Im{G[k]}
Re{G[k]}
Two N-point DFT's

4
2
1
-1
100
200
Frequency index k
Single N-point DFT
-2
300
1
Im{G[k]}
Re{G[k]}
2
1
0
100
200
Frequency index k
Single N-point DFT
300
100
200
Frequency index k
300
0
-1
100
200
Frequency index k
-2
300
32
Two N-point DFT's
Two N-point DFT's

4
2
Im{H[k]}
Re{H[k]}
-2
100
200
Freq index k
Single N-point DFT
-4
300
100
200
Freq index k
Single N-point DFT
300
100
200
Freq index k
300
4
2
Im{H[k]}
Re{H[k]}
0
-2
100
200
Freq index k
-4
300
In the first example, the length is

N=256. The sequence g[n] is real and even and the sequence h[n] is real but has no
inherent symmetry. The DFT G[k] is real and even to within numerical roundoff; it is
important to realize in the plots that Im{G[k]} is zero to within floating point precision.
In the second example, the length is N=128. The sequence g[n] is a sample function of
an IID stochastic process uniformly distributed between +0.8 and -0.8. The sequence
h[n] is the product of a deterministic ramp and a deterministic decaying exponential.
In the third example, the length is N=300. The sequence g[n] is an alternating and
decaying exponential, while sequence h[n] is the product of a monotonically decaying
exponential and a cosine.
In all three of these examples, we see that the results obtained by computing G[k] and
H[k] with a single N-point DFT or directly with two N-point DFTs are identical up to
numerical roundoff.
33
The MATLAB program to compute the 2N-point DFT of a length-2N real sequence using a
single N-point DFT and compare the result by computing directly the 2N-point DFT is shown
below:
% Program Q3_25A
% Use a single N-point complex-valued DFT to compute the 2N-point
%
DFT of a 2N-point real-valued sequence.
%
clf;
%Initialize constants
N = 128;
% length of the complex-valued DFT
TwoN = 2*N;
% length of the real-valued sequence
W2N = exp(-i*pi/N);
k = [0:TwoN-1];
% create 2N-point signal v[n]
v = (-0.7.^k);
% create two N-point signals
g = downsample(v,2);
% g[n] = v[2n]
h = downsample(v,2,1); % h[n] = v[2n+1]
% N-point complex-valued composite signal x[n]
x = g + i*h;
% Use one N-point complex DFT to compute simultaneously
%
G[k] and H[k] as in Q3.24.
XF = fft(x);
% N-point complex DFT of x[n]
% DFT's G[k] and H[k] derived from composite DFT X[k]
XFstar = conj(XF);
XFstarmod = [XFstar(1) fliplr(XFstar(2:N))];
GF = 0.5*(XF + XFstarmod);
HF = -i*0.5*(XF - XFstarmod);
% 2N-point DFT V[k]
VF = [GF GF] + (W2N.^k).*[HF HF];
% For Comparison, compute directly the 2N-point DFT V[k]
VF2 = fft(v);
% Print Sanity Check
abs(max(VF-VF2))
% plot real and imag parts of V[k] computed by complex N-point DFT
subplot(2,2,1);
plot(k,real(VF));grid;
title('Complex N-point DFT');
ylabel('Re\{V[k]\}');
subplot(2,2,2);
plot(k,imag(VF));grid;
title('Complex N-point DFT');
ylabel('Im\{V[k]\}');
% plot real and imag parts of V[k] computed by 2N-point DFT
subplot(2,2,3);
plot(k,real(VF2));grid;
title('Real 2N-point DFT');
ylabel('Re\{V[k]\}');
subplot(2,2,4);
plot(k,imag(VF2));grid;
title('Real 2N-point DFT');
ylabel('Im\{V[k]\}');
Q3.25
34
The DFTs generated by running the program for sequences of different lengths 2N are shown
below:
Complex N-point DFT
2
-1
1
Im{V[k]}
Re{V[k]}
Complex N-point DFT

0
-2
-3
-1
100
200
Frequency index k
Real 2N-point DFT
-2
300
-1
1
Im{V[k]}
Re{V[k]}
-4
-2
-3
-4
100
200
Frequency index k
Real 2N-point DFT
300
100
200
Frequency index k
300
0
-1
100
200
Frequency index k
-2
300
Complex N-point DFT
Complex N-point DFT

4
2
Im{V[k]}
Re{V[k]}
-2
500
Frequency index k
Real 2N-point DFT
-4
1000
500
Frequency index k
Real 2N-point DFT
1000
500
Frequency index k
1000
4
2
Im{V[k]}
Re{V[k]}
0
-2
500
Frequency index k
-4
1000
35
Complex N-point DFT

40
20
20
Im{V[k]}
Re{V[k]}
Complex N-point DFT

40
0
-20
-20
200
400
Frequency index k
Real 2N-point DFT
-40
600
40
40
20
20
Im{V[k]}
Re{V[k]}
-40
0
-20
-40
200
400
Frequency index k
Real 2N-point DFT
600
200
400
Frequency index k
600
0
-20
200
400
Frequency index k
-40
600
In the first example, the length of

v[n] is 256 and V[k]is computed using a complex valued 128-point DFT. The signal
is v[n] (0.7) n . In the second example, the length of v[n] is 1000 and V[k]is
computed using a complex valued 500-point DFT.
The signal is
n
v[n] 0.8 cos n 25 . In the third example, the length of v[n] is 600 and V[k]is
computed using a complex valued 600-point DFT. In each case, the results obtained by
the 2N-point real DFT and by the N-point complex DFT are identical up to numerical
roundoff.
Project 3.4
DFT Properties
Answers:
Q3.26
The purpose of the command rem in the function circshift is
rem(x,y) is the
remainder after x is divided by y.

Q3.27
The input sequence x is circularly shifted

left by M positions. If M > 0, then circshift removes the leftmost M elements from
the vector x and appends them on the right side of the remaining elements to obtain the
circularly shifted sequence. If If M < 0, then circshift first complements M by the
length of x, i.e., the rightmost length(x)-M samples are removed from x and appended
on the right of the remaining M samples to obtain the circularly shifted sequence.
The function circshift operates as follows:
36
This is the binary relational

NOT EQUAL operator. A ~= B returns the value 1 if A and B are unequal and the
value 0 if A and B are equal.
Q3.28
The purpose of the operator ~= in the function circonv is
Q3.29
The function circonv operates as follows:
Q3.30
program statements for labeling each plot being generated by the program is given below:
The input is two vectors x1 and x2 of equal

length L. To understand how circonv works, it is useful to think in terms of the
periodic extension of x2. Let x2p be the infinite-length periodic extension of x2.
Conceptually, the routine time reverses x2p and sets x2tr equal to elements 1 through L
of the time reversed version of x2p. Elements 1 through L of the output vector y are
then obtained by taking the inner product between x1 and a length L vector sh obtained
by circularly shifting right the time reversed vector x2tr. For the output sample y[n], 1
n L, the amount of the right circular shift is n-1 positions.
% Program P3_7B
% Illustration of Circular Shift of a Sequence
clf;
% initialize shift amount M
M = 6;
% initialize sequence a to be shifted
a = [0 1 2 3 4 5 6 7 8 9];
b = circshift(a,M); % perform the circular shift
L = length(a)-1;
% plot the original sequence a and the circularly shifted sequence b
n = 0:L;
subplot(2,1,1);
stem(n,a);axis([0,L,min(a),max(a)]);
title('Original Sequence');
xlabel('time index n');
ylabel('a[n]');
subplot(2,1,2);
stem(n,b);axis([0,L,min(a),max(a)]);
title(['Sequence Obtained by Circularly Shifting by ',num2str(M),'
Samples']);
xlabel('time index n');
ylabel('b[n]');
The parameter determining the amount of time-shifting is -
The circular shift

actually implemented is rem(M,length(a)) positions left, which is equivalent to
circularly shifting by M positions (more than once around) and also to shifting left by
M the periodic extension of the sequence.
If the amount of time-shift is greater than the sequence length then
37
Original Sequence
8
a[n]
6
4
2
0
4
5
6
7
8
time index n
Sequence Obtained by Circularly Shifting by 12 Samples
8
6
b[n]
Q3.31
4
2
0
4
5
time index n
Here, the length of the sequence is

10 samples and we have M=12. This may be interpreted alternatively as a circular shift
left by 12 positions (more than once around), as a circular shift left by 12-10 = 2
positions, or as a linear shift left by 2 or by 12 of the periodic extension of the
sequence.
38
Q3.32
program statements for labeling each plot being generated by the program is given below:
% Program P3_8B
% Circular Time-Shifting Property of DFT
clf;
x = [0 2 4 6 8 10 12 14 16]; % original sequence x
N = length(x)-1; n = 0:N;
% time index vector
% set y equal to the circular shift left of x
y = circshift(x,5);
XF = fft(x);
% DFT of x
YF = fft(y);
% DFT of y
subplot(2,2,1);
% plot the spectral magnitudes of the original and shifted sequences
stem(n,abs(XF));grid;
title('Magnitude of DFT of Original Sequence');
ylabel('|X[k]|');
subplot(2,2,2);
stem(n,abs(YF));grid;
title('Magnitude of DFT of Circularly Shifted Sequence');
ylabel('|Y[k]|');
% plot the spectral phases of the original and shifted sequences
subplot(2,2,3);
stem(n,angle(XF));grid;
title('Phase of DFT of Original Sequence');
ylabel('arg(X[k])');
subplot(2,2,4);
stem(n,angle(YF));grid;
title('Phase of DFT of Circularly Shifted Sequence');
ylabel('arg(Y[k])');
The amount of time-shift is

Q3.33
- hard coded in this program at 5 samples to the left.
39
Magnitude of DFT of Original Sequence

Magnitude of DFT of Circularly Shifted Sequence
80
80
60
|Y[k]|
|X[k]|
60
40
20
0
20
2
4
6
8
Frequency index k
Phase of DFT of Original Sequence
4
2
4
6
8
Frequency index k
Phase of DFT of Circularly Shifted Sequence
4
2
arg(Y[k])
arg(X[k])
2
0
-2
-4
40
0
-2
2
4
6
Frequency index k
-4
2
4
6
Frequency index k
The length of the sequence is N=8

and the time shift is an advance by five samples to the left. The phase term introduced
by this time shift is WNkn0 WN k 5 e jk 10 / 8 e jk 5 / 4 . This is a substantial shift that
dramatically increases the slope of the spectral phase. Whereas the original phase
function has only one branch cut, there are five branch cuts in the spectral phase of the
shifted signal.
Q3.34
The plots generated by running the modified program for the following two different amounts of
time-shifts, with the amount of shift indicated, are shown below:
40
M=2
80
80
60
|Y[k]|
|X[k]|
60
40
20
0
20
2
4
6
8
Frequency index k
4
2
4
6
8
Frequency index k
4
2
arg(Y[k])
arg(X[k])
2
0
-2
-4
40
0
-2
2
4
6
Frequency index k
-4
2
4
6
Frequency index k
M=-2
80
80
60
|Y[k]|
|X[k]|
60
40
20
0
20
2
4
6
8
Frequency index k
4
2
4
6
8
Frequency index k
4
2
arg(Y[k])
arg(X[k])
2
0
-2
-4
40
0
-2
2
4
6
Frequency index k
-4
41
2
4
6
Frequency index k
From these plots we make the following observations: In all cases, the spectral magnitude
is not affected by the shift. For the first example, the time shift is a circular shift left by
2 samples. This introduces an increased slope to the spectral phase that is significantly
less than what we saw in Q3.33. In the second example, the shift is circular shift right
by 2 samples (M=-2). This cancels the positive slope seen in the spectral phase of the
original sequences and results in a moderate negative slope.
The plots generated by running the modified program for the following two different sequences
of different lengths, with the lengths indicated, are shown below:
Length = 24
Magnitude of DFT of Original SequenceMagnitude of DFT of Circularly Shifted Sequence
150
|Y[k]|
|X[k]|
150
100
100
50
0
50
10
20
Frequency index k
4
10
20
Frequency index k
4
arg(Y[k])
arg(X[k])
Q3.35
0
-2
2
0
-2
10
15
20
Frequency index k
42
10
15
20
Frequency index k
Length = 16
Magnitude of DFT of Original SequenceMagnitude of DFT of Circularly Shifted Sequence
150
|Y[k]|
|X[k]|
150
100
100
50
0
50
10
20
Frequency index k
4
2
10
20
Frequency index k
4
arg(Y[k])
arg(X[k])
0
-2
2
0
-2
10
15
20
Frequency index k
10
15
20
Frequency index k
In the first example, the sequence is

real and periodically (circularly) even, so the phase takes only the two values zero and
. The shift is a circular shift to the right by 2 (M=-2), which is seen to induce a
negative slope to the phase. In the second example, the signal is given by (0.75)n and
the shift is again M=-2 which again introduces a negative slope in the phase.
Q3.36
A copy of Program P3_9 is given below along with the plots generated by running this program:
% Program P3_9
% Circular Convolution Property of DFT
g1 = [1 2 3 4 5 6]; g2 = [1 -2 3 3 -2 1];
ycir = circonv(g1,g2);
disp('Result of circular convolution = ');disp(ycir)
G1 = fft(g1); G2 = fft(g2);
yc = real(ifft(G1.*G2));
disp('Result of IDFT of the DFT products = ');disp(yc)
Result of circular convolution =

12 28 14 0 16 14
Result of IDFT of the DFT products =
12 28 14 0 16 14
43
The DFT of a circular convolution is
the pointwise products of the DFTs.

Q3.37
Program P3_9 was run again for the following two different sets of equal-length sequences:
The plots generated are shown below:
Result of circular convolution =

9
2 -7
Result of IDFT of the DFT products =

9.0000 2.0000 -7.0000 6.0000 9.0000 4.0000
The circular convolution property of
the DFT seems to hold.

Q3.38
A copy of Program P3_10 is given below along with the plots generated by running this
program:
% Program P3_10
% Linear Convolution via Circular Convolution
g1 = [1 2 3 4 5];g2 = [2 2 0 1 1];
g1e = [g1 zeros(1,length(g2)-1)];
ylin = circonv(g1e,g2e);
disp('Linear convolution via circular convolution = ');disp(ylin);
y = conv(g1, g2);
disp('Direct linear convolution = ');disp(y)
Linear convolution via circular convolution =

2
6 10 15 21 15
Direct linear convolution =

2
6 10 15 21 15
zero padding to the appropriate

length does indeed make it possible to implement linear convolution using circular
convolution.
Q3.39
Program P3_10 was run again for the following two different sets of sequences of unequal
lengths:
g1 = [3 1 4 1 5 9 2];
g2 = [1 1 1 0 0];
44
g1 = [5 4 3 2 1 0];
g2 = [-2 1 2 3 4];
The plots generated are shown below:

3 4 8 6 10 15 16 11 2 0 0
3

8 6 10 15 16 11 2

-10 -3 8 22 38 30 20 11 4 0
-10 -3

8 22 38 30 20 11
You can implement the linear

convolution of two sequences by zero padding them to the sum of their lengths less one
and then invoking circular convolution on the zero padded sequences.
Q3.40
The MATLAB program to develop the linear convolution of two sequences via the DFT of each is
given below:
% Program Q3_40
% Linear Convolution via Circular Convolution
g1 = [1 2 3 4 5];
g2 = [2 2 0 1 1];
G1EF = fft(g1e);
G2EF = fft(g2e);
ylin = real(ifft(G1EF.*G2EF));
disp('Linear convolution via DFT = ');disp(ylin);
The plots generated by running this program for the sequences of Q3.38 are shown below:
Linear convolution via DFT =

2.0000 6.0000 10.0000 15.0000 21.0000 15.0000 7.0000 9.0000
Q3.38; in other words, it works as advertised.
45
The result is the same as before in
The plots generated by running this program for the sequences of Q3.39 are shown below:
Columns 1 through 9
3.0000
4.0000
8.0000
6.0000 10.0000 15.0000 16.0000 11.0000
2.0000
Columns 10 through 11
0.0000
0.0000

Columns 1 through 9
-10.0000 -3.0000
8.0000 22.0000 38.0000 30.0000 20.0000 11.0000
4.0000
Column 10
-0.0000
The results are the same as those

that were obtained before when the DFT was not used.
Q3.41
% Program P3_11
% Relations between the DFTs of the Periodic Even
% and Odd Parts of a Real Sequence
x = [1 2 4 2 6 32 6 4 2 zeros(1,247)];
x1 = [x(1) x(256:-1:2)];
xe = 0.5 *(x + x1);
XF = fft(x);
XEF = fft(xe);
clf;
k = 0:255;
subplot(2,2,1);
plot(k/128,real(XF)); grid;
title('Re(DFT\{x[n]\})');
subplot(2,2,2);
plot(k/128,imag(XF)); grid;
title('Im(DFT\{x[n]\})');
subplot(2,2,3);
plot(k/128,real(XEF)); grid;
xlabel('Time index n');ylabel('Amplitude');
title('Re(DFT\{x_{e}[n]\})');
subplot(2,2,4);
plot(k/128,imag(XEF)); grid;
title('Im(DFT\{x_{e}[n]\})');
46
The relation between the sequence x1[n] and x[n] is
x1[n] is a periodically time
reversed version of x[n].

The plots generated by running Program P3_11 are given below:
Re(DFT{x[n]})
Im(DFT{x[n]})
100
50
Amplitude
Amplitude
100
-50
0.5
1.5
50
0
-50
-100
Re(DFT{x e[n]})
Amplitude
0.5
1
1.5
Time index n
0.5
-14
50
-50
0
x 10
100
Amplitude
Q3.42
1.5
Im(DFT{x e[n]})
0.5
0
-0.5
0.5
1
1.5
Time index n
The imaginary part of XEF is equal to zero to within floating point precision. This result can be
explained as follows: The real part of the transform of x[n] is the transform of the
periodically even part of x[n]. Therefore, the DFT of the periodically even part of x[n]
has a real part that is precisely the real part of X[k] and an imaginary part that is zero.
47
Q3.43
The required modifications to Program P3_11 to verify the relation between the DFT of the
periodic odd part and the imaginary part of XEF are given below along with the plots
generated by running this program:
% Program P3_11B
% Relations between the DFTs of the Periodic Even
% and Odd Parts of a Real Sequence
x = [1 2 4 2 6 32 6 4 2 zeros(1,247)];
x1 = [x(1) x(256:-1:2)];
xo = 0.5 *(x - x1);
XF = fft(x);
XOF = fft(xo);
clf;
k = 0:255;
subplot(2,2,1);
plot(k/128,real(XF)); grid;
title('Re(DFT\{x[n]\})');
subplot(2,2,2);
plot(k/128,imag(XF)); grid;
title('Im(DFT\{x[n]\})');
subplot(2,2,3);
plot(k/128,real(XOF)); grid;
title('Re(DFT\{x_{o}[n]\})');
subplot(2,2,4);
plot(k/128,imag(XOF)); grid;
title('Im(DFT\{x_{o}[n]\})');
48
Re(DFT{x[n]})
Im(DFT{x[n]})
100
50
Amplitude
Amplitude
100
-50
0.5
-14
1.5
-50
0.5
Re(DFT{x o[n]})
1.5
Im(DFT{x o[n]})
100
0.5
0
-0.5
-100
Amplitude
Amplitude
x 10
50
0.5
1
1.5
Time index n
50
0
-50
-100
0.5
1
1.5
Time index n
The DFT of the periodically odd

part of x[n] is precisely the imaginary part of the DFT of x[n]. Therefore, the DFT of
the periodically odd part of x[n] has a real part that is zero to within floating point
precision and an imaginary part that is precisely the imaginary part of the DFT of x[n].
Q3.44
% Program P3_12
% Parseval's Relation
x = [(1:128) (128:-1:1)];
XF = fft(x);
a = sum(x.*x)
b = round(sum(abs(XF).^2)/256)
The values for a and b we get by running this program are
a =1414528
b =1414528
49
Q3.45
The required modifications to Program P3_11 are given below:
% Program P3_12B
% Parseval's Relation
x = [(1:128) (128:-1:1)];
XF = fft(x);
a = sum(x.*x)
b = round(sum(XF.*conj(XF))/256)
3.3
z-TRANSFORM
Project 3.5
Analysis of z-Transforms
Answers:
The frequency response of the z-transform obtained using Program P3_1 is plotted below:
Real part of H(ej )
Amplitude
0.6
0.4
0.2
0
-0.2
0.1
0.2
0.3
0.4
0.5
/
0.6
0.7
0.8
0.9
0.7
0.8
0.9

0
Amplitude
Q3.46
-0.1
-0.2
-0.3
-0.4
0.1
0.2
0.3
0.4
0.5
50
0.6
Amplitude
0.8
0.6
0.4
0.2
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.8
0.9
/
Phase in radians
0
-1
-2
-3
-4
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Q3.47
The MATLAB program to compute and display the poles and zeros, to compute and display the
factored form, and to generate the pole-zero plot of a rational z-transform is given below:
% Program Q3_47
% Given numerator and denominator coefficient vectors for G(z),
%
- compute and display poles and zeros
%
- compute and display factored form of G(z)
%
- generate pole-zero plot
% NOTE: the lab book says to use tf2zp. For a rational function
%
in z^-1, it's better to use tf2zpk.
clf;
% initialize
num = [2 5 9 5 3];
den = [5 45 2 1 1];
% compute poles and zeros and display
[z p k] = tf2zpk(num,den);
disp('Zeros:');
disp(z);
disp('Poles:');
disp(p);
input('Hit <return> to continue...');
% compute and display factored form of G(z)
[sos k] = zp2sos(z,p,k)
input('Hit <return> to continue...');
% generate pole-zero plot
zplane(z,p);
51
Using this program we obtain the following results on the z-transform G(z) of Q3.46:
Zeros:
-1.0000 + 1.4142i
-1.0000 - 1.4142i
-0.2500 + 0.6614i
-0.2500 - 0.6614i
Poles:
-8.9576
-0.2718
0.1147 + 0.2627i
0.1147 - 0.2627i
sos =
1.0000 2.0000 3.0000 1.0000 9.2293 2.4344
1.0000 0.5000 0.5000 1.0000 -0.2293 0.0822
k=
0.4000
1 2 z 1 3z 2
1 0.5 z 1 0.5 z 2
G ( z ) 0.4
1 9.2293z 1 2.4344 z 2 1 0.2293z 1 0.0822 z 2
4
3
Imaginary Part
2
1
0
-1
-2
-3
-4
-9
-8
-7
-6
-5
-4
-3
Real Part
52
-2
-1
Q3.48
From the pole-zero plot generated in Question Q3.47, the number of regions of convergence
(ROC) of G(z) are - FOUR. note: the magnitude of the complex conjugate poles
inside the unit circle is 0.2866.

All possible ROCs of this z-transform are sketched below:
R1 : | z |
0.2718
(left-sided, not stable)
R2 : 0.2718 | z |
0.2866
(two-sided, not stable)
R3 : 0.2866 | z |
8.9576
(two-sided, stable)
R4 : | z |
8.9576
(right-sided, not stable)
You cannot tell if the DTFT exists

from the pole zero plot alone. In order to know this, the region of convergence must be
specified. The DTFT does exist for the sequence obtained by using the ROC R3 shown
above. This would be a stable system with a two-sided impulse response.
From the pole-zero plot it can be seen that the DTFT
Q3.49
The MATLAB program to compute and display the rational z-transform from its zeros, poles
and gain constant is given below:
% Program Q3_49
% Given the poles and zeros of G(z), compute and display the rational
% z-transform.
clf;
% initialize
z = [0.3 2.5 -0.2+i*0.4 -0.2-i*0.4]';
p = [0.5 -0.75 0.6+i*0.7 0.6-i*0.7]';
k = 3.9;
% find numerator and denominator polynomial coefficients
[num den] = zp2tf(z,p,k)
The rational form of a z-transform with the given poles, zeros, and gain is found to be
num =
3.9000 -9.3600 -0.6630 -1.0140 0.5850
den =
1.0000 -0.9500 0.1750 0.6625 -0.3187
G( z)
3.9 9.36 z 1 0.663z 2 1.014 z 3 0.585 z 4

1 0.95 z 1 0.175 z 2 0.6625 z 3 0.3187 z 4
53
Project 3.6
Inverse z-Transform
Answers:
Q3.50
The MATLAB program to compute the first L samples of the inverse of a rational z-transform is
given below:
% Program Q3_50
% find and plot the first L samples of the impulse response, where
% the parameter L is input by the user.
%
clf;
% initialize
num = [2 5 9 5 3];
den = [5 45 2 1 1];
% Query user for parameter L
L = input('Enter the number of samples L: ');
% find impulse response
[g t] = impz(num,den,L);
%plot the impulse response
stem(t,g);
title(['First ',num2str(L),' samples of impulse response']);
xlabel('Time Index n');
ylabel('h[n]');
The plot of the first 50 samples of the inverse of G(z) of Q3.46 obtained using this program is
sketched below:
45
First 50 samples of impulse response
x 10
0
-2
-4
h[n]
-6
-8
-10
-12
-14
-16
10
15
20
25
30
Time Index n
54
35
40
45
50
Q3.51
The MATLAB program to determine the partial-fraction expansion of a rational z-transform is

given below:
% Program Q3_51
% find and plot the first L samples of the impulse response, where
% the parameter L is input by the user.
%
clf;
% initialize
num = [2 5 9 5 3];
den = [5 45 2 1 1];
% partial fraction expansion
[r p k] = residuez(num,den)
The partial-fraction expansion of G(z) of Q3.46 obtained using this program is shown below:
r=
0.3109
-1.0254 - 0.3547i
-1.0254 + 0.3547i
-0.8601
p=
-8.9576
0.1147 + 0.2627i
0.1147 - 0.2627i
-0.2718
k=
3
G( z)
0.3109
1.0254 0.3547 j
1.0254 0.3547 j
0.8601
3
1
1
1
1 8.9576 z
1 (0.1147 0.2627 j ) z
1 (0.1147 0.2627 j ) z
1 0.2718 z 1
From the above partial-fraction expansion we arrive at the inverse z-transform g[n] as shown below:
Three of the terms are straightforward to invert from the z-transform table on page 305 of the
book:
0.3109
1 8.9576 z 1
0.3109(8.9576) n u[n]
0.8601
1 0.2718 z 1
0.8601(0.2718) n u[n]
55

3 3[ n]
There remains the term

1.0254 0.3547 j
1.0254 0.3547 j
.
1
1 (0.1147 0.2627 j ) z
1 (0.1147 0.2627 j ) z 1
(1.1)
Let a 1.0254 0.3547 j and b 0.1147 0.2627 j. To save writing, let aR Re[a ] ,
aI Im[a ] , bR Re[b] , and bI Im[b]. The term (1.1) may then be written as
a
a
.
1
1 bz
1 b z 1
(1.2)
After some algebra, (1.2) can be simplified to

2aR 2aR bR 2aI bI z 1
a
a
1 bz 1 1 b z 1
1 2bR z 1 | b |2 z 2
2aR 2aR bR z 1
2aI bI z 1
1 2bR z 1 | b |2 z 2 1 2bR z 1 | b |2 z 2
2aR
(1.3)
1 bR z 1
bI z 1
2
a
.
I
1 2bR z 1 | b |2 z 2
1 2bR z 1 | b |2 z 2
Making the associations r | b | and 0 arg b , we obtain bR r cos(0 ) and

bI r sin(0 ) , whereupon (1.3) may be written as
1 r cos(0 ) z 1
r sin(0 ) z 1
a
a
2
a
2
a
. (1.4)
R
I
1 bz 1 1 b z 1
1 2r cos(0 ) z 1 r 2 z 2
1 2r cos(0 ) z 1 r 2 z 2
These are the last two entries in the z-transform table on page 281 of the text. Therefore,
the inverse z-transform of (1.4) is given by
2aR r n cos(0 n)u[n] 2aI r n sin(0 n)u[n].
(1.5)
Plugging back into (1.5) the definitions of a , b , r , and 0 , we have for the inverse ztransform of the term (1.1)
2.0508(0.2866)n cos(1.1592n)u[n] 0.7094(0.2866)n sin(1.1592n)u[n].
56
All together, the required inverse z-transform is given by

g[n] 0.3109(8.9576) n u[n] 2.0508(0.2866) n cos(1.1592n)u[n]
0.7094(0.2866) n sin(1.1592n)u[n] 0.8601(0.2718) n u[n] 3[n]
Date:
24 October 2006
Signature: HAVLICEK
57

HW 06 Sol

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

HW 06 Sol

Hochgeladen von

Copyright:

Verfügbare Formate

Name: SOLUTION (Havlicek)

DISCRETE-TIME FOURIER TRANSFORM

A copy of Program P3_1 is given below:

The expression of the DTFT being evaluated in Program P3_1 is - H e

The plots generated by running Program P3_1 are shown below:

Imaginary part of H(ej )

Magnitude Spectrum |H(ej )|

Phase Spectrum arg[H(ej )]

The DTFT is a _periodic_ function of .

The real part is 2 periodic and EVEN SYMMETRIC.

The imaginary part is 2 periodic and ODD SYMMETRIC.

The magnitude is 2 periodic and EVEN SYMMETRIC.

The phase is 2 periodic and ODD SYMMETRIC.

Imaginary part of H(ej )

Magnitude Spectrum |H(ej )|

Phase Spectrum arg[H(ej )]

The DTFT is a _periodic__ function of .

Imaginary part of H(ej )

Magnitude Spectrum |H(ej )|

The DTFT is a _periodic_ function of .

Only the last paragraph of the code needs to be changed to:

The parameter controlling the amount of time-shift is - D

Magnitude Spectrum of Original Sequence

Phase Spectrum of Original Sequence

Phase Spectrum of Time-Shifted Sequence

Magnitude Spectrum of Time-Shifted Sequence

Magnitude Spectrum of Original Sequence

Phase Spectrum of Original Sequence

Phase Spectrum of Time-Shifted Sequence

1. Length 4, time shift D=3.

Magnitude Spectrum of Time-Shifted Sequence

Magnitude Spectrum of Original Sequence

Phase Spectrum of Original Sequence

Phase Spectrum of Time-Shifted Sequence

Phase Spectrum of Original Sequence

Phase Spectrum of Time-Shifted Sequence

Magnitude Spectrum of Time-Shifted Sequence

Magnitude Spectrum of Original Sequence

The parameter controlling the amount of frequency-shift is - wo.

Magnitude Spectrum of Original Sequence

Magnitude Spectrum of Frequency-Shifted Sequence

Phase Spectrum of Original Sequence

Phase Spectrum of Frequency-Shifted Sequence

From these plots we make the following observations:

Both the magnitude and phase

Magnitude Spectrum of Original Sequence

Magnitude Spectrum of Frequency-Shifted Sequence

Phase Spectrum of Original Sequence

Phase Spectrum of Frequency-Shifted Sequence

From these plots we make the following observations:

In this case, the magnitude and

1. Length 4, frequency shift wo = .

Magnitude Spectrum of Frequency-Shifted Sequence

Magnitude Spectrum of Original Sequence

Phase Spectrum of Original Sequence

Phase Spectrum of Frequency-Shifted Sequence

Phase Spectrum of Original Sequence

Phase Spectrum of Frequency-Shifted Sequence

Magnitude Spectrum of Frequency-Shifted Sequence

Magnitude Spectrum of Original Sequence

Magnitude Spectrum of Convolved Sequence

Sum of Phase Spectra

Phase Spectrum of Convolved Sequence